Portugal Released Amália — the First National AI Model for European Portuguese

Portugal released Amália in July 2026 — the first national AI language model created specifically for European Portuguese. The country with a population of just over 10 million chose the most transparent approach: not only the model weights were published, but also the training data on which it was built.

Why European Portuguese requires a separate model

For English or Spanish speakers, the difference between European and Brazilian Portuguese may seem insignificant. For users and developers — it is fundamental. It is not just about accent: these are different lexical norms, syntactic constructions, idioms, and spelling rules, established after the 2009 reform.

Brazil, with its 200-million population, generates incomparably more digital content than Portugal. Large international language models are trained on what is most abundant on the internet — and European Portuguese loses to Brazilian Portuguese many times over in this race. For Portuguese users, this manifests concretely: models mix forms of address, incorrectly process official administrative language, poorly understand local idioms and legal constructions. Amália was developed specifically to close this gap.

What the government made public

Key facts of the project:

• Amália — Portugal's first official national language AI model
• Trained specifically on European Portuguese, not Brazilian
• Open access: model weights and training dataset published
• Initiator — the Portuguese state; population of the country — approximately 10 million people
• Release date — July 2026

Publishing training data is rare even among academic open-source projects. It allows independent researchers and developers not only to use the finished model, but also to understand what it was trained on, fine-tune it for specific tasks, and verify quality without a "black box." Government agencies, universities, startups, and medical institutions gain access without licensing barriers.

Amália and European AI sovereignty

The release fits into the broader European discussion of how smaller countries can avoid complete dependence on American and Chinese language models. The EU AI Act, adopted in 2024, encourages member states to develop their own technological competencies — and Portugal implements this principle literally.

For small and medium European languages — Dutch, Czech, Greek, Hungarian, Finnish — the gap in AI processing quality compared to English is particularly noticeable. Large laboratories are not economically motivated to invest proportional resources in languages with 10–50 million speakers: the market does not create quality models where there is insufficient training data volume or commercial attractiveness. State initiatives like Amália fill exactly this niche — and an open license turns the model into a public good that accumulates and improves through community effort.

What this means

Portugal has demonstrated a working precedent: a country with a population of 10 million is capable of independently creating and releasing a language model for its own language variant. If this approach proves reproducible — and open data and weights allow others to replicate it — it could become a template for the Netherlands, Czech Republic, Greece, and other European states whose languages remain on the periphery of global AI systems.